Proposals to ban artificial intelligence models are increasingly viewed by technology policy experts as insufficient tools for managing the complex risks associated with generative AI. While the call for prohibitions often stems from concerns over safety, data privacy, and intellectual property, a growing consensus among regulators and industry stakeholders suggests that an actionable agenda—focused on transparent testing, safe release protocols, and stable infrastructure—is necessary to address the technology’s rapid evolution. Relying on blanket bans often overlooks the reality that AI development is decentralized, global, and moving faster than legislative processes can typically accommodate.
According to the White House Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence, signed in October 2023, the federal government has pivoted toward establishing rigorous safety standards, including requirements for developers of powerful AI systems to share their safety test results with the government. This shift toward “actionable agendas” reflects a departure from the idea that the technology can be halted or contained through simple prohibition. Instead, the focus has moved toward risk management frameworks that allow for innovation while imposing guardrails on how models are deployed in critical sectors.
The Limitations of Prohibitive Policy
The primary critique of banning AI models is the “whack-a-mole” effect, where restrictive measures in one jurisdiction do not stop development elsewhere. As noted in the European Union’s AI Act, which was formally adopted in May 2024, the strategy is to implement a risk-based classification system rather than a total ban on the technology itself. The EU legislation categorizes AI systems by risk level, ranging from “unacceptable” (which are banned) to “high risk” (which require strict compliance with transparency and oversight rules). This nuanced approach acknowledges that AI has both societal benefits and potential harms, making a blanket ban an impractical and overly blunt instrument.
Legal scholars and policy analysts argue that focusing on bans can also stifle domestic competitiveness. By restricting the development of frontier models within a specific country, regulators might inadvertently push research and development into jurisdictions with fewer oversight requirements. The National Institute of Standards and Technology (NIST) AI Risk Management Framework emphasizes that the goal of policy should be the creation of “trustworthy” systems. This requires continuous monitoring, evaluation, and iteration, rather than a one-time approval or a permanent prohibition.
Building an Actionable Agenda
An effective policy framework requires three core components: technical transparency, standardized safety testing, and clear accountability for developers. The United Kingdom government’s approach to AI regulation, outlined in its policy papers, prioritizes a pro-innovation stance that relies on existing regulators to enforce safety within their specific domains—such as finance, medicine, and transport—rather than creating a single, rigid, and potentially obsolete AI-specific ban.
Actionable agendas are increasingly being defined by the following requirements:
- Red-teaming: Mandating that developers perform adversarial testing to identify vulnerabilities before a model is released to the public.
- Reporting requirements: Ensuring that companies disclose the compute power used to train models, as highlighted in the Biden-Harris administration’s specific reporting mandates for large-scale models.
- Interoperable Standards: Developing international consensus on what constitutes “safe” AI to ensure that safety protocols are consistent across borders.
Economic and Safety Implications
The debate over how to regulate AI is also an economic one. According to a report by the International Monetary Fund published in early 2024, AI is expected to impact nearly 40% of global employment. Because the technology is deeply integrated into the global economy, a ban would be economically destabilizing. Instead, policymakers are being urged to focus on “stable access”—ensuring that the benefits of AI, such as productivity gains in healthcare and education, are available to the public while minimizing the risks of bias, misinformation, and cyberattacks.
The move toward “stable access” also addresses the issue of market concentration. When policies are too restrictive or costly to comply with, only the largest, most well-funded firms can afford to develop and deploy AI models. This creates a barrier to entry for smaller startups and academic researchers. By focusing on risk management rather than prohibition, governments can maintain a more competitive and diverse ecosystem of AI developers, which in turn fosters safer, more scrutinized technology.
Next Steps in Global Governance
The international community is currently moving toward the next phase of AI oversight. The U.S. Department of State and other international bodies are engaged in ongoing dialogues regarding the “Bletchley Declaration,” which was signed in November 2023 by 28 countries, including the U.S. and China. This agreement commits signatories to a shared understanding of the risks posed by frontier AI and the need for international cooperation on safety research.
The next major checkpoint for global AI policy will be the implementation phase of the EU AI Act, with various compliance deadlines beginning in 2025 and 2026. In the United States, federal agencies are continuing to update their internal guidance in accordance with the 2023 Executive Order. Readers interested in tracking these developments can monitor official updates from the U.S. Artificial Intelligence Safety Institute, which serves as a central hub for federal efforts to test and evaluate AI systems. As policies shift from abstract concerns to practical implementation, the focus will remain on balancing the rapid pace of innovation with the necessity of public safety.
What are your thoughts on current AI regulatory trends? Share your perspective in the comments below.
Keep reading